Improved image deblurring based on combination of generalized Gaussian distribution and nonlocal-means
改进的广义高斯分布与非局部均值图像去模糊

In order to improve the traditional algorithm can't reserve the detail information of image defects, and obtain better image to the blurring effect, this paper proposed a new image deblurring algorithm based on generalized Gaussian distribution and NL-means. Firstly, it proposed the blurred image with the wavelet transform, and then estimated scale parameter and shape parameter of generalized Gaussian distribution model with OMLE and classical Newton-Raphson algorithm. It used these two parameters were used to improve initial image weights calculating method which was singly judged by decay rate of exponential function and restricted with only one algorithm. Experimental results on several typical images indicate that the proposed algorithm gives superior effects to initial NL-means and has a good application prospects.